Research And Implementation Of Mobile O2O Recommendation System Based On LBS

With the widespread growth of mobile devices,mobile applications are provided to users by enterprises in 020 business area,and millions of users with mobile application produced a lot of rating data with location context information.Analyzing users' rating data history and providing potential interesting recommendation list of 020 services for users are important measures in sales promotion.Meanwhile,consideration of users' and services' features such as mobility in mobile 020 recommendation to improve efficiency is the core mission in recommendation system construction.Current personalized recommendation methods cannot take advantages of the users' location context information,and the personalized recommendation results produced cannot meet the individual needs of the user's location,resulting in a large number of unused calculation.Meanwhile,serious matrix sparsity problem is faced by traditional collaborative filtering technology,and the recommending efficiency is greatly affected.Therefore,a location-based clustering collaborative filtering algorithm was raised by this article for mobile 020 service recommendation and take full advantage of LBS technology and user's location context information,and a mobile 020 recommendation system based on LBS was designed and implemented.Location-based clustering collaborative filtering algorithm mainly based on off-line calculation,by clustering users' rating data with location information and decomposing the rating matrix,reducing the sparsity of the rating matrix,finally recommends with collaborative filtering.And the location crossing algorithm is raised to solve the problem of insufficient recommendation when users moved to unfamiliar locations.The improved collaborative filtering algorithm is based on the distributed machine-learning platform Apache Mahout,and the produced recommendation results are applied to the real-time recommendation module,to expose interfaces to third party systems through RESTful interfaces.Finally,the real user data experiment shows the location-based clustering collaborative filtering algorithm has better efficiency than traditional collaborative filtering algorithm with appropriate chosen of the clustering number,and the improved clustering collaborative filtering algorithm can accommodate the scene with large ranges of location and rating data,to fully meet the users' real-time and personalized requirements.